RESEARCH Open Access

Low shifts in salinity determined assemblyprocesses and network stability ofmicroeukaryotic plankton communities in asubtropical urban reservoirYuanyuan Mo1,2, Feng Peng1, Xiaofei Gao1,2, Peng Xiao1, Ramiro Logares3, Erik Jeppesen4,5,6,7, Kexin Ren1,Yuanyuan Xue1 and Jun Yang1*

Abstract

Background: Freshwater salinization may result in significant changes of microbial community composition anddiversity, with implications for ecosystem processes and function. Earlier research has revealed the importance oflarge shifts in salinity on microbial physiology and ecology, whereas studies on the effects of smaller or narrowershifts in salinity on the microeukaryotic community in inland waters are scarce. Our aim was to unveil communityassembly mechanisms and the stability of microeukaryotic plankton networks at low shifts in salinity.

Results: Here, we analyzed a high-resolution time series of plankton data from an urban reservoir in subtropicalChina over 13 consecutive months following one periodic salinity change ranging from 0 to 6.1‰. We found that(1) salinity increase altered the community composition and led to a significant decrease of plankton diversity, (2)salinity change influenced microeukaryotic plankton community assembly primarily by regulating the deterministic-stochastic balance, with deterministic processes becoming more important with increased salinity, and (3) coreplankton subnetwork robustness was higher at low-salinity levels, while the satellite subnetworks had greaterrobustness at the medium-/high-salinity levels. Our results suggest that the influence of salinity, rather thansuccessional time, is an important driving force for shaping microeukaryotic plankton community dynamics.

Conclusions: Our findings demonstrate that at low salinities, even small increases in salinity are sufficient to exert aselective pressure to reduce the microeukaryotic plankton diversity and alter community assembly mechanism andnetwork stability. Our results provide new insights into plankton ecology of inland urban waters and the impacts ofsalinity change in the assembly of microbiotas and network architecture.

Keywords: Subtropical reservoir, Microeukaryotic plankton, Community ecology, Network stability, Core taxa,Satellite taxa, Salinity, Deterministic processes, Stochastic processes

© The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you giveappropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate ifchanges were made. The images or other third party material in this article are included in the article's Creative Commonslicence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commonslicence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtainpermission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to thedata made available in this article, unless otherwise stated in a credit line to the data.

* Correspondence: jyang@iue.ac.cn1Aquatic Ecohealth Group, Fujian Key Laboratory of Watershed Ecology, KeyLaboratory of Urban Environment and Health, Institute of UrbanEnvironment, Chinese Academy of Sciences, Xiamen 361021, ChinaFull list of author information is available at the end of the article

Mo et al. Microbiome (2021) 9:128 https://doi.org/10.1186/s40168-021-01079-w

IntroductionFreshwater salinization (i.e., increasing salt concentra-tion) is becoming an extensive global environmentalproblem potentially caused by saltwater intrusion,urbanization, and climate change, especially in semi-aridand arid climate zones [1–5]. Saltwater intrusion canjeopardize drinking water resources and infrastructuresuch as tubing systems, which may suffer from greaterwear [6]. Severe salinization has also detrimental influ-ences on freshwater ecosystems [7], such as lethal effects(loss of species diversity) and fitness reduction of fresh-water organisms (sublethal effects) in high-salinityconditions. For example, suppression of growth [8], de-creased feeding efficiency [9], and increased deformitiesin frogs [10] have been observed with increasing salinity.Similarly, aquatic micro-organisms tend to be negativelyimpacted by high salinity in hyper-saline lakes, asreflected by decreased microbial diversity [11].Microeukaryotic plankton communities are key com-

ponents of aquatic ecosystems and play a significant eco-logical role in controlling food web structuring andcarbon flow through photosynthesis [12, 13]. Under-standing the ecological processes determining the com-munity assembly of these microorganisms is central tothe field of community ecology [14, 15]. Deterministicand stochastic processes explain the assembly of micro-bial communities [15, 16]. Deterministic processes in-volve both biotic and abiotic factors, namely interspeciesinteractions (e.g., competition, predation, mutualism,and tradeoff) and environmental filtering (e.g., salinity,pH, temperature), which together shape communitycomposition [17, 18]. Stochastic processes consider thatall species are ecologically equivalent, and include ran-dom birth, death, dispersal, extinction, and speciation,which also affect community assembly [19, 20]. Micro-eukaryotic community assembly may be strongly affectedby stochastic processes in rivers and oceans [21, 22].However, there is evidence that deterministic processesmay be the dominant ecological mechanisms determin-ing the community assembly of microeukaryotic plank-ton at determined spatial scales [23]. In most cases, bothdeterministic and stochastic processes can jointly shapemicroeukaryotic plankton communities [24, 25]. Previ-ous studies have revealed that broad salinity change isone of the most significant environmental variablesshaping microbial community structure in aquatic andterrestrial ecosystems (see Additional file 1: Table S1).Yet, few studies have investigated the influences of lowshifts in salinity on the community assembly of microeu-karyotic plankton in inland freshwaters.How environmental changes, such as salinity and

drought disturbance, affect microbial assembly and thebalance between deterministic and stochastic processesremains unclear. As both processes are already

governing the microeukaryotic plankton community as-sembly in natural ecosystems, they will likely be pro-moted or limited by environmental changes across timeand space. A previous study found that fungal commu-nity stochasticity did not increase when drought stresswas relieved and attributed this to strong deterministicselection imposed by the host in the sorghum system[26]. Another study revealed that a strong selection pres-sure was imposed by salinity on the soil microbial com-munity in desert ecosystems, resulting in dominance ofdeterministic processes under high-salinity conditions[27]. Importantly, salinity-driven selection is regarded asa major factor affecting the balance of assembly mecha-nisms in soil bacterial communities [28]. Evidently, tem-poral dynamics are often associated with changes inenvironmental conditions, which complicate our under-standing of the mechanisms underpinning communityassembly [29]. However, few studies have explored thechange in deterministic processes relative to the changesin stochastic counterparts induced by low shifts in salin-ity in inland freshwaters.Further, environmental disturbances likely destabilize

microbial co-occurrence networks [30]. In natural eco-systems, most microeukaryotic plankton communitiesconsist of a few core taxa with high abundances and ahuge number of satellite taxa with low abundances [31,32]. A previous study hypothesized that in macro-organisms core taxa are mainly influenced by selection,whereas satellite taxa are mostly affected by dispersallimitation [33]. Such partition has been valuable forcomprehending the ecological processes shaping macro-organism communities [34, 35], thereby contributing toa better understanding of ecosystem functions [36].There is increasing evidence indicating that communityresponses to environmental disturbances can be affectedby ecological network characteristics [37]. For example,weak interactions and co-exclusion lead to more stabletemporal ecological networks [38]. Despite a surge ofnew and insightful network analyses in ecology, signifi-cant knowledge gaps remain regarding how microeukar-yotic plankton community stability responds to lowshifts in salinity.Estuarine or brackish waters are ideal systems for eluci-

dating microbial dynamics with changing salinity [14, 39,40]. However, unlike estuaries and saline lakes, our studyarea, the Xinglinwan Reservoir, is a freshwater urban res-ervoir in a rapidly urbanizing area (Jimei District of Xia-men City) of southeast China, subjected to periodicsalinization. The resident population in Jimei district ofXiamen City jumped from 148,000 in 2000 to about 580,000 in 2010, and then to 1037,000 in 2020. After the con-struction of the Xingji seawall in 1979, the XinglinwanReservoir gradually evolved from the natural bay to thepresent enclosed water body [41]. The reservoir was

Mo et al. Microbiome (2021) 9:128 Page 2 of 17

disconnected from the ocean by a sluicegate [41], whichmay be the main factor causing the salinity changes ob-served in this water body. In addition, terrestrial input,water discharge from surrounding human activities, andprecipitation may also have contributed to salinity changes[42]. The Xinglinwan Reservoir is the most importantwater body in the Jimei District and plays an importantrole for the landscape, water storage, and flood control inXiamen City. Yet, it is not clear how low shifts in salinityin the reservoir affect the diversity and community assem-bly of microeukaryotic plankton as well as their co-occurrence patterns. Here, we examined the dynamics ofmicroeukaryotic plankton communities in this subtropical

urban reservoir using 18S rRNA gene sequencing basedon high-frequency sampling (daily to weekly) over a 13-month period. To facilitate comparison between differentsalinity levels at low shifts in salinity, we artificially dividedthe samples into three salinity categories: low-salinity (0–0.2‰), medium-salinity (0.2–2‰), and high-salinity (2–6.1‰) conditions (Fig. 1a, b).Two hypotheses were tested: (i) low shifts in salinity at

low salinities would significantly affect the compositionand diversity of the microeukaryotic plankton communi-ties in freshwaters through a progressive increase in de-terministic processes and a decrease in stochasticcounterparts; (ii) the co-occurrence network stability of

Fig. 1 Sampling sites and principal coordinates analysis of the microeukaryotic plankton community. a Location of the three sampling sites inXinglinwan Reservoir, Xiamen City, Southeast China. Water samples were taken from stations C, L, and G. b Principal coordinates analysis (PCoA)of microeukaryotic plankton community composition at station G in the Xinglinwan Reservoir. Each circle represents one sample and is color-coded according to time (month) and sized according to salinity

Mo et al. Microbiome (2021) 9:128 Page 3 of 17

core and satellite plankton subcommunities would bedifferent with low shifts in salinity.

ResultsTemporal dynamics of environmental variablesA suite of environmental factors characterizing the Xin-glinwan Reservoir are visualized in Figures S1 and S2 inAdditional file 1. Compared with station C, several simi-lar temporal tendencies of the physical and chemicalfactors were found at stations L and G. Salinity, elec-trical conductivity (EC), and total nitrogen (TN) showedalmost synchronous changes over time, while salinityand precipitation/temperature demonstrated the oppos-ite trend (Additional file 1: Figure S2). Furthermore,opposite trends were observed for salinity and precipita-tion between 2016 and 2018, and Spearman’s correlationindicated that salinity and precipitation exhibited a sig-nificant and negative correlation from 2016 to 2018(Additional file 1: Figure S3).

Composition and temporal dynamics of microeukaryoticplankton communitiesThe total number of microeukaryotic OTUs was 19,952in the reservoir at 97% similarity level, which included618 core OTUs and 18,056 satellite OTUs (Additionalfile 1: Table S2, Additional file 1: Figure S4a, b). Con-tinuous fluctuations were observed in each planktonsupergroup of the microeukaryotic community, and themost abundant OTUs were assigned to the groups ofAlveolata, Archaeplastida, Cryptista, Opisthokonta, andStramenopiles (Additional file 1: Figure S5a). In addition,the core taxa were more abundant but less diverse thanthe satellite taxa at phylum level (Additional file 1:Figure S5b). The absolute abundance of the microeukar-yotic communities ranged from 4.19 × 109 to 4.04 ×1010 copies/L at station C, while the absolute abundanceat stations L and G varied from 4.21 × 109 to 7.92 × 1010

copies/L and from 4.29 × 108 to 6.63 × 1011 copies/L, re-spectively. The fluctuations in absolute abundance ofmicroeukaryotes in the time series were larger than thatof bacterioplankton (Additional file 1: Figure S6a, b).Changes in the ratio of microeukaryotic 18S rRNA geneto bacterial 16S rRNA gene were observed across thetime series (Additional file 1: Figure S6c), however theabundance of microeukaryotic 18S rRNA gene exhibited asignificant positive correlation with bacterial 16S rRNAgene at stations C, L, and G, respectively (Additional file 1:Figure S6d).Non-metric multi-dimensional scaling (NMDS)

ordination showed a significant segregation of whole-community microeukaryotic OTUs (based on 97% se-quence similarity) between the three salinity levels atstations C, L, and G, whereas the microeukaryotic com-munities at stations C, L, and G exhibited a significant

aggregation at low-salinity conditions, indicating no sig-nificant spatial difference in the effects of salinity oncommunity composition. Whole-community ASVs(amplicon sequence variants) results showed almostidentical pattern as those based on OTUs (ρ = 0.997,P = 0.001, Additional file 1: Figure S7). Therefore, weselected only one station, station G, for high-frequency monitoring during 13 consecutive monthsto explore the shifts of microeukaryotic OTUs (Add-itional file 1: Figure S7).

The importance of salinity in structuring planktoncommunitiesMicroeukaryotic plankton communities were separatedbased on salinity or time (month) at station G (Fig. 1).Our results showed that salinity exhibited the strongestcorrelations with all, core, and satellite microeukaryoticcommunity composition, followed by alpha- and beta-diversities of bacterioplankton (Fig. 2). Other significantenvironmental variables were water temperature, pH,dissolved oxygen, chlorophyll-a, turbidity, electrical con-ductivity, total carbon, total organic carbon, total nitro-gen, ammonium nitrogen, nitrate nitrogen, nitritenitrogen, total phosphorus, and phosphate phosphorus(Fig. 2). In addition, the all community Bray-Curtis dis-similarity exhibited a stronger correlation with salinity(R = 0.774, P < 0.01) than with time (R = 0.658, P < 0.01for absolute time; R = 0.646, P < 0.01 for annual cycletime) (Additional file 1: Figure S8). The Mantel and par-tial Mantel results also revealed that both salinity andtime significantly explained the change in microeukaryo-tic community composition (including all, core, and sat-ellite taxa; P < 0.01), whereas salinity had a greaterinfluence on microeukaryotic communities than time(Additional file 1: Table S3). Furthermore, the effects ofsalinity on alpha-diversity of all, core, and satellite plank-ton were stronger than time or the interaction of salinityand time (Table 1).

Microeukaryotic plankton community composition anddiversity along salinity gradientNon-metric multi-dimensional scaling (NMDS) ordin-ation and ANOSIM tests showed that the compositionof the microeukaryotic plankton communities (all, core,and satellite taxa) differed significantly between low-,medium-, and high-salinity levels (R = 0.623, P = 0.001for all taxa; R = 0.536, P = 0.001 for core taxa; and R =0.633, P = 0.001 for satellite taxa, Fig. 3a). Further, allcore OTUs were shared among the three salinity levels,while the proportion of shared OTUs (27.3%) was muchlower for the satellite taxa (Fig. 3b). Shannon-Wiener di-versity decreased with increasing salinity for all, core,and satellite plankton taxa (Fig. 3c), with Chlorophytaand Ochrophyta being the dominant groups. The

Mo et al. Microbiome (2021) 9:128 Page 4 of 17

relative abundance of Chlorophyta and Ochrophyta in-creased and decreased, respectively, with increasing sal-inity, especially for all and core taxa (Additional file 1:Figure S9).

Relative importance of deterministic and stochasticprocesses along salinity gradientThe relationship between the occurrence frequency ofOTUs and their relative abundance was well describedby the neutral community model (Fig. 4a). The relativecontribution of stochastic processes decreased graduallywith increasing salinity, explaining 78.5%, 58.5%, and48.3% of the community variance for the low, medium,and high-salinity levels, respectively. The same patternwas observed in the succession time series, reflectingthat the contribution of stochastic processes to theplankton community was low when salinity was high

(Additional file 1: Figure S10). Further, all microeukaryo-tic plankton communities exhibited significantly widerniche breadths at low salinity than at medium-/high-sal-inity levels (Fig. 4b). The average niche breadth wassignificantly higher for core than for the satellitesubcommunities (31.5 for core taxa, 10.5 for satellitetaxa; P < 0.001) (Additional file 1: Figure S11). More im-portantly, the C-score showed that standardized effectsize (SES) increased with increasing salinity, indicatingthe enhanced importance of deterministic processes forthe plankton assemblage (Fig. 4c).

Co-occurrence networks and stability of microeukaryoticplankton communities along salinity gradientA metacommunity co-occurrence network was con-structed based on all datasets from station G, and threesubnetworks along three salinity levels (low, medium,

Fig. 2 Abiotic and biotic drivers of microeukaryotic plankton community composition. Pairwise comparisons of environmental and biotic factorsare shown at the upper-right, with a color gradient representing Spearman’s correlation coefficients. Microeukaryotic plankton communitycomposition was correlated to each environmental or biotic factor by partial Mantel tests. The line width represents the partial Mantel’s r statisticfor the corresponding correlation, and line color means that significances are tested based on 999 permutations. WT, water temperature; DO,dissolved oxygen; Chl-a, chlorophyll-a; EC, electrical conductivity; ORP, oxidation-reduction potential; TC, total carbon; TOC, total organic carbon;TN, total nitrogen; NH4-N, ammonium nitrogen; NO3-N, nitrate nitrogen; NO2-N, nitrite nitrogen; TP, total phosphorus; PO4-P, phosphatephosphorus; Note that the precipitation data are the 7-day accumulation before the sampling day, and the wind represents daily average windspeed. B_richness, bacterial OTU number; B_SW, bacterial Shannon-Wiener index; B_NMDS1, bacterial NMDS ordination axis 1; B_NMDS2, bacterialNMDS ordination axis 2. Note that only significant correlations are shown for simplicity

Mo et al. Microbiome (2021) 9:128 Page 5 of 17

and high) were analyzed for all, core, and satellitetaxa (Additional file 1: Table S4), respectively. Thetopological properties of the networks varied signifi-cantly with salinity. For example, both nodes (i.e.,OTUs) and edges (i.e., significant links or correlationsbetween OTUs) constituting the all and core net-works decreased with increasing salinity (that is, thenetworks became smaller with increasing salinity).Also, the satellite networks consisted of differentnodes linked by different edges along salinity levels,316 nodes being linked by 755 edges, 290 nodes by1316 edges, and 166 nodes by 582 edges in low-,medium-, and high-salinity levels, respectively (Add-itional file 1: Table S4). Further, all network degreesfollowed a power-law distribution rather than a Pois-son distribution, suggesting that the network structureexhibited a scale-free and non-random distribution(Additional file 1: Table S4). The observed network

parameters (average clustering coefficient, averagepath length, and modularity index) were all largerthan those of their respective Erdös-Réyni randomnetworks, indicating “small-world” properties andmodular structure (Additional file 1: Table S4). Basedon the all network, the nodes with the top threehighest degrees were OTU_3 (Fungi), OTU_37(Chlorophyta), and OTU_98 (Ochrophyta), being po-tential keystone species. Both OTU_3 and OTU_37belonged to the core microeukaryotic plankton (Add-itional file 1: Figure S12a). Random forest (RF) ana-lysis indicated that the beta-diversity of satellitesubcommunities exhibited a stronger relationship withthe multi-nutrient cycling index than the core sub-communities (Additional file 1: Figure S12b).The plankton community network was clearly di-

vided into six major modules that accounted for82.4% of the whole network (Fig. 5a). For example,microeukaryotic plankton communities were domi-nated by taxa preferring medium salinity in module I,by taxa preferring high salinity in modules II, IV, andVI, and taxa preferring low salinity in modules IIIand V. Further, we found that the contribution ofOchrophyta, Perkinsea, and Ciliophora was higher inmodules representing low salinity. However, in themodules corresponding to medium and high-salinitylevels, Chlorophyta exhibited the highest degree ofcentrality (Additional file 1: Table S5). This indicatesthat Ochrophyta, Perkinsea, and Ciliophora play a keyrole in maintaining taxa coexistence in low-salinitynetwork, whereas Chlorophyta are more important inmaintaining coexistence in the medium and high-salinity networks because nodes with a higher degreeof centrality are more important in maintaining taxacoexistence in networks.Finally, we compared network stability between differ-

ent salinity levels with varying microeukaryotic planktonsubnetwork structure. Compared with core taxa, the dis-similarity of subnetworks was larger between the lowand high-salinity levels for both all and satellite taxa(Additional file 1: Table S6). In fact, the community dis-similarities between the groups were always higher thanthose within groups for the different salinity levels, andsatellite subcommunities showed a higher dissimilaritywhen compared against all and core taxa. When com-paring the core subnetwork stability among the threesalinity levels, the natural connectivity at low salinitywas greater than that at medium and high-salinity levels.However, natural connectivity at low salinity was muchlower than at medium or high salinity for satellite plank-ton subnetworks (Fig. 5b). This indicates greater coresubnetwork robustness at low-salinity levels, whereas thesatellite subnetwork had greater robustness at themedium and high-salinity levels.

Table 1 Two-way ANOVA showing the effects of time andsalinity on the alpha-diversity of microeukaryotic planktoncommunities

All Core Satellite

F P F P F P

Time (n = 13)

Richness 2.978 0.001 1.957 0.037 3.337 0.000

ACE 3.201 0.001 1.654 0.090 3.555 0.000

Chao 1 3.152 0.001 1.622 0.098 3.479 0.000

Shannon-Wiener 2.431 0.008 1.990 0.033 0.931 0.520

Simpson 0.870 0.580 0.246 0.995 0.564 0.866

Pielou’s evenness 2.086 0.025 4.640 0.000 0.445 0.941

Salinity (n = 3)

Richness 16.767 0.000 20.729 0.000 11.986 0.000

ACE 16.867 0.000 19.355 0.000 13.914 0.000

Chao 1 16.366 0.000 20.885 0.000 12.492 0.000

Shannon-Wiener 15.628 0.000 8.396 0.000 15.216 0.000

Simpson 3.429 0.036 0.359 0.699 0.684 0.507

Pielou’s evenness 10.873 0.000 14.707 0.000 1.615 0.204

Time × salinity

Richness 3.991 0.002 2.550 0.033 3.084 0.013

ACE 3.951 0.003 2.194 0.061 3.456 0.006

Chao 1 3.869 0.003 2.054 0.078 3.223 0.010

Shannon-Wiener 4.528 0.001 2.379 0.044 2.294 0.051

Simpson 0.773 0.571 0.343 0.885 0.916 0.474

Pielou’s evenness 3.676 0.004 4.691 0.001 0.487 0.785

Boldface means significance at P < 0.05 levelTime includes 13 successional monthsSalinity includes low, medium, and high salinity levelsAll all microeukaryotic plankton communities, Core core microeukaryoticplankton subcommunities, Satellite satellite microeukaryoticplankton subcommunities

Mo et al. Microbiome (2021) 9:128 Page 6 of 17

DiscussionLow salinity increase triggers plankton communitycomposition change and diversity lossIn this study, we obtained new data on microeukaryoticplankton community dynamics at low shifts in salinityin an urban freshwater habitat based on high-resolutionsampling. We found that salinity variation may inducecompositional changes and diversity loss in the all,core, and satellite plankton communities. PlanktonShannon-Wiener diversity decreased with increasingsalinity for all, core and satellite taxa, emphasizing thatmicroeukaryotic plankton communities were not resili-ent when subjected to salinity variation (Fig. 3c). Plank-ton sensitivity to salinity may reflect the increase inextracellular osmolarity with increasing salinity [43],meaning that microeukaryotic plankton communitiesthat are unable to adapt to osmotic stress are likely todie or become less active, leading to a decrease inalpha-diversity of the communities. This is consistentwith a former study [44], showing that mycoplanktonalpha-diversity was higher during low-saline periods in

coastal ecosystems. However, a whole ecosystem ma-nipulation experiment with freshwater rock pools withthree salinity levels (3, 6, and 12 ‰) did not find anynegative effect on bacterial alpha-diversity of a limitedsalinity increase [45]. The contrasting results may berelated to the different microbial groups studied (bac-terial community in previous study and the microeu-karyotic plankton community in this study). Lowsalinity increase may not be sufficient to filter out orsuppress many bacterial taxa but sufficed to stronglysuppress many microeukaryotic plankton species, per-haps reflecting the high abundance, fast growth rates,and rapid evolutionary adaptation by bacteria [46].

Salinity mediates the assembly processes ofmicroeukaryotic plankton communitiesOur study indicates that low shifts in salinity in freshwa-ters have an important influence on the assembly of allmicroeukaryotic plankton communities, primarily by af-fecting the balance between deterministic and stochasticprocesses. The community variation explained by

Fig. 3 Community structuring of microeukaryotic plankton across the salinity gradient at station G in Xinglinwan Reservoir. a Non-metricmultidimensional scaling (NMDS) ordination based on Bray-Curtis dissimilarity showing the variation of microeukaryotic plankton communitiesacross three salinity levels. Significant level of all, core, and satellite taxa is P = 0.001. b Venn diagram showing the numbers of unique and sharedOTUs between three different salinity levels. c Shannon-Wiener index along the salinity level at station G. Different letters indicate significantdifference at P < 0.05 according to Tukey’s post-hoc test. All, all microeukaryotic plankton communities at station G; Core, core microeukaryoticplankton subcommunities at station G; Satellite, satellite microeukaryotic plankton subcommunities at station G

Mo et al. Microbiome (2021) 9:128 Page 7 of 17

stochastic processes decreased from 78.5% at low-salinity level to 48.3% at high-salinity level (Fig. 4a). Inaddition, the microeukaryotic plankton communitiesshowed wider niche breadths under low salinity than atthe medium-/high-salinity conditions (Fig. 4b), implyingthat the community assembly was more strongly influ-enced by deterministic processes at high salinity, likelybecause deterministic processes tend to have a strongereffect on habitat specialists with a narrow niche breadththan on generalists with a wide niche breadth [23, 47].Further, C-score results showed that the value of stan-dardized effect size (SES) increased with increasing salin-ity, also indicating that the community assembly wasmore strongly influenced by deterministic processes withincreasing salinity (Fig. 4c). There are several possible

explanations. First, increased allochthony may increasestochastic processes in the wet season based on otherenvironmental factors that do not impose strong selec-tion [48]. Specifically, precipitation events can increasethe freshwater input to the Xinglinwan Reservoir, ac-companied by a decrease in salinity. Meanwhile, in-creased precipitation may wash the micro-organismsfrom surrounding environmental systems (the soil orsediment, watershed, and air) into the reservoir, leadingto an enhanced immigration rate and higher diversity ofmicrobial community, which again increases stochasti-city at low salinity. This agrees with the higher microeu-karyotic diversity in low-salinity periods, followed bymedium and high-salinity periods (Fig. 3). Second, atlow salinity, freshwater microeukaryotes may be exposed

Fig. 4 Ecological processes shaping the microeukaryotic plankton community assembly at station G in Xinglinwan Reservoir. a The predictedoccurrence frequencies for low, medium, and high salinity representing microeukaryotic plankton communities from low, medium, and highsalinity periods in Xinglinwan Reservoir. The solid blue line is the best fit to the neutral community model (NCM), and the dashed blue lineindicates 95% confidence intervals around the NCM prediction. OTUs that occur more or less frequently than predicted by the NCM are shown ingreen and red, respectively. R2 represents the fit to this model. b Comparison of mean habitat niche breadth for all taxa among low, medium,and high salinity levels (different letters indicate significant difference at the P < 0.05 level using Tukey’s post hoc test). c C-score metric using nullmodels. The values of observed C-score (C-scoreobs) > simulated C-score (C-scoresim) indicate non-random co-occurrence patterns. Standardizedeffect size <− 2 and > 2 represent aggregation and segregation, respectively

Mo et al. Microbiome (2021) 9:128 Page 8 of 17

to lower physiological stress, so that they can grow andreproduce more freely, resulting in dominance of sto-chastic processes (e.g., birth, death, and dispersal events)in community assembly, while a strong selective pressuremay be exerted on the freshwater microeukaryoticplankton when salinity increases [27]. Microbes thatgrow well at high salinity have developed “salt-in” and “salt-out” strategies to adjust the cytoplasm to osmotic pressure[43]. “Salt-in” often involves intake of ions (e.g., K+ andCl–), while “salt-out” maintains a low-intracellular ion con-centration through pumping out inorganic ions and accu-mulating compatible solutes (e.g., sucrose, glycerol, and

glycin) to exclude salt from the cell and thus ensure os-motic balance [43]. This would expectedly result in a highercontribution of deterministic processes to the communityassembly with increasing salinity. Third, in low-salinity eco-systems with less environmental heterogeneity or with lesscompetitive interactions between environmental generalists,the stochastic assembly mechanism is likely to overrule de-terministic processes [49]. Our results suggest that the de-gree to which deterministic vs. stochastic processes shapethe microeukaryotic plankton community in this reservoiris determined more by shifts in salinity rather than by sea-sonality (Table 1; Additional file 1: Table S3).

Fig. 5 Network modules and stability for microeukaryotic plankton OTUs at station G. a Network revealing the modular associations amongmicroeukaryotic plankton OTUs (left). Relative abundance of microeukaryotic OTUs from major modules along the three different salinity levels(right). A connection indicates a strong (SparCC |r| > 0.6) and significant (P < 0.01) correlation. The size of each microeukaryotic OTU (node) isproportional to the number of connections (i.e., degree). b Network stability for all, core and satellite taxa at different salinity levels (i.e., low,medium, and high), respectively. The robustness of all, core, and satellite networks under different salinity conditions in Xinglinwan Reservoir.Insert: overview of pairwise community dissimilarity of microeukaryotic plankton communities at three different salinity levels. Statistical analysis isnon-parametric Mann-Whitney U test. ***P < 0.001. All, all microeukaryotic plankton communities; Core, core microeukaryotic planktonsubcommunities; Satellite, satellite microeukaryotic plankton subcommunities

Mo et al. Microbiome (2021) 9:128 Page 9 of 17

Co-occurrence network stability of core and satellitemicroeukaryotic plankton shaped by salinityAlthough the stability of plankton communities can beused to infer ecosystem functioning [12], it is unclearhow plankton co-occurrence networks in an urban res-ervoir respond to disturbances such as small changes insalinity. In particular, the stability of core or satelliteplankton subnetworks is largely unexplored [30, 50].Our results revealed that the six network modules corre-sponded well with the three salinity levels (Fig. 5a). Thisindicates that the modular structure or property of allmicroeukaryotic plankton communities was sensitive tochanges in salinity. Environmental heterogeneity couldinduce microbial modularity [51], which helps explainwhy the six modules became dominant at different salin-ity levels. Modularity can reflect competitive/synergisticrelationships and niche differentiation, yielding non-random patterns of network structure, which ultimatelyincreases the complexity of ecological networks [52].The network modules may act as indicators of importantecological processes following the disturbance of salinitychanges.Based on the plankton network modules, we found

that core OTUs play important roles in maintaining net-work stability (Additional file 1: Figure S12a). Similar re-sults have been described for microbiota in metropolitandrinking water [53], agricultural soils [54], and activatedsludge ecosystems [55]. Interestingly, the contribution ofthe beta-diversity of the satellite subcommunity to theaquatic ecosystem multi-nutrient cycling index wasgreater than that of the core subcommunity (Additionalfile 1: Figure S12b). This suggests that satellite taxa alsoplay key roles in maintaining ecosystem functions. Ourresults indicate that microbial diversity affects the multi-functionality in aquatic ecosystems.Core plankton subnetworks were less sensitive to sal-

inity changes than satellite taxa at low-salinity variability(Fig. 5b). At low salinity, the satellite subnetworks wereless stable, while core subnetworks showed a higher sta-bility. At higher salinity, the opposite pattern was foundfor the stability of core and satellite plankton subnet-works, which may be explained as follows. First, thenetwork topological parameter ‘average clustering coeffi-cient’ was the highest for core subnetworks and the low-est for satellite sub-networks at low salinity (Additionalfile 1: Table S4). This implies a higher complexity of thecore subnetwork and a lower complexity of the satellitesubnetwork at low salinity. High complexity networksnormally tend to have greater stability due to networkbuffering [56], so core microeukaryotic plankton subnet-works were likely more stable and satellite subnetworksmore unstable at low salinity. Second, our results indi-cate that core taxa had wider niche breadths than satel-lite taxa (Additional file 1: Figure S11), meaning that

they, compared with satellite taxa, can adapt to a widerange of environmental niches [57]. Consequently, coreco-occurrence subnetworks exhibited strong resistanceas salinity increased, while satellite subnetworks areprone to be affected by slight disturbances of salinity.This pattern could be closely associated with planktondiversity and satellite taxa richness. In other words, sal-inity affects plankton diversity and satellite taxa richnessand, as a consequence, network structure and connectiv-ity. Thus, it is to be expected that core and satelliteplankton exhibited similar patterns in alpha- and beta-diversities but different ecosystem stability patternsalong the salinity gradient. It is, therefore, important todistinguish between core and satellite taxa of microeu-karyotic plankton when assessing community stabilityover time and future threats to ecosystem function andservices in response to environmental disturbance.

ConclusionWe propose a conceptual framework to describe themicroeukaryotic plankton community responses to lowshifts in salinity in inland freshwaters (Fig. 6). Low in-creases in salinity decrease the alpha-diversity of micro-eukaryotic plankton communities. All, core, and satelliteplankton communities show the strongest relationshipwith salinity, followed by temperature and bacterial di-versity or community due to microbial interactions orsynchronous dynamics. Importantly, our results help un-raveling the mechanisms affecting the balance betweenthe deterministic and stochastic assembly of microeukar-yotic plankton with changes in salinity. The potentialkeystone species in the all network belong to the coretaxa, and the beta-diversity of satellite subcommunitiessignificantly affects multi-nutrient cycling, implying thatcore and satellite OTUs play important but differentroles in maintaining ecological function. In addition,core plankton networks are more stable in low-salinityenvironments, whereas satellite networks are more stablein the medium-/high-salinity environments. Given that alow increase in salinity in this freshwater reservoir sig-nificantly influenced the plankton community, manage-ment and protection require better knowledge of theresponse of the plankton community to salinity changesand their interactions with other human or natural dis-turbances [3, 4, 58], when evaluating, modeling, and pre-dicting salinity effects on coastal urban freshwaterecosystems.

MethodsStudy station, sampling, and environmental informationSurface water samples were collected in Xinglinwan Res-ervoir, Xiamen City, Fujian Province, Southeast China,at three stations (station C: 24° 36′ 53′′ N, 118° 03′ 11′′E; station L: 24° 36′ 21′′ N, 118° 03′ 37′′ E; station G:

Mo et al. Microbiome (2021) 9:128 Page 10 of 17

24° 36′ 09′′ N, 118° 03′ 59′′ E) (Fig. 1a). Specifically,samples were collected approximately daily from August12, 2016 to August 30, 2016 at station C (12 samples),daily or twice a week from August 12 to September 20in 2016 at station L (22 samples), and approximatelydaily from August 12 to September 20, 2016 and thentwice a week from September 23, 2016 to August 18,2017 at station G (116 samples). Each water sample wasdivided into two subsamples: one for microeukaryoticplankton analyses and the other for water chemistry ana-lyses. About 500 mL of surface water (upper 50 cm) wasfiltered through a 200-μm mesh to remove larger parti-cles and then filtered through 0.22-μm pore-size poly-carbonate membrane filters (47-mm diameter, Millipore,Billerica, MA, USA) to collect the microeukaryotic cellswithin 60-min. The filters were then stored at – 80 °Cuntil further analysis.In addition, 18 environmental variables were measured

or collected (Additional file 1: Figure S1 and S2). Watertemperature, pH, dissolved oxygen, turbidity, electricalconductivity, salinity, and oxidation-reduction potential(ORP) were measured in situ with a Hydrolab DS5multiparameter water quality analyzer (Hach Company,Loveland, CO, USA). Chl-a concentrations were quanti-fied by PHYTO-PAM Phytoplankton Analyzer (Heinz

Walz GmbH, Eichenring, Germany). Total carbon (TC),total organic carbon (TOC), total nitrogen (TN), ammo-nium nitrogen (NH4-N), nitrate nitrogen (NO3-N), ni-trite nitrogen (NO2-N), total phosphorus (TP), andphosphate phosphorus (PO4-P) were measured accord-ing to the standard methods described in our previousstudy [59]. Precipitation and daily average wind speeddata were downloaded from the Xiamen MeteorologicalBureau. The precipitation data consisted of a 7-dayaccumulation before the sampling day. To study therelationship between salinity and precipitation, dataon daily precipitation were collected and 3-year salin-ity data were measured from 2016 to 2018 (Add-itional file 1: Figure S3).

DNA extraction, PCR, and Illumina sequencingThe total DNA was extracted from the filters using theFastDNA SPIN Kit and the FastPrep Instrument (MPBiomedicals, Santa Ana, CA, USA) following the manu-facturer’s instructions. For the microeukaryotic planktoncommunity, the V9 region of eukaryotic 18S rRNA genewas amplified using the primer pair 1380F and 1510R[60]. The 30-μL PCR reaction included 15-μL of PhusionHigh-Fidelity PCR Master Mix (New England Biolabs,Beverly, MA, USA), 0.2 μM of each primer, and 10 ng of

Fig. 6 Conceptual models of microeukaryotic plankton diversity, community assembly processes, and network stability driven by low shiftsin salinity

Mo et al. Microbiome (2021) 9:128 Page 11 of 17

community DNA. The PCR reactions included initial de-naturation at 98 °C for 1 min, followed by 30 cycles of10 s at 98 °C, 50 °C for 60 s, and 72 °C for 30 s. At theend of the amplification, the amplicons were subjectedto a final 10 min extension at 72 °C. The PCR productsfrom triplicate reactions per sample were pooled andgel-purified. All libraries were sequenced on the IlluminaHiSeq platform (Illumina Inc., San Diego, CA, USA)using a paired-end (2 × 150 bp) approach.To explore whether bacterial communities had an im-

pact on the microeukaryotic plankton community, theV3-V4 hypervariable regions of the 16S rRNA gene wereamplified, purified, and quantified following our previousprocedure [61]. The triplicate PCR products were pooledtogether, and sequencing was performed on the IlluminaMiSeq platform (Illumina, Inc., San Diego, CA, USA)using 2 × 250 bp paired-end sequencing approach.

BioinformaticsPairs of reads from both the 18S rRNA and 16S rRNAgene were processed using VSEARCH v.2.14.1 [62].Quality check and sequence merge were conductedusing MOTHUR v.1.39.5 [63], and the filtered reads(“sequences”) were then processed as unique sequencesusing “minuniquesize 8” parameter in VSEARCH. Theunoise3 algorithm was used to discard chimeras and as-sign operational taxonomic units (OTUs) at a 97% se-quence similarity threshold in USEARCH v11 [64].Subsequently, for microeukaryotic plankton, representa-tive sequences from each OTU were taxonomically clas-sified using an 80% confidence threshold against theProtist Ribosomal Reference (PR2) reference sequencedatabase [65]. To make the taxonomic classificationmore user friendly and portable, the taxonomic assign-ments were adjusted to be in accordance with the taxo-nomic reference of eukaryotes [66, 67]. To minimizeinclusion of sequencing errors, OTUs present in < 5samples with < 10 sequences were excluded from thedownstream analyses. After the OTU table was gener-ated, we randomly rarefied a subset of 146,973 se-quences from each of the 150 samples to standardize thesequencing effort using MOTHUR v.1.39.5 [63]. The146,973 sequences were selected because they repre-sented the sample with the lowest sequence numberfrom all the samples. Finally, the total dataset retained19,952 OTUs and 22,045,950 sequences at 97% similaritythreshold. OTU numbers of unclassified Eukaryotaaccounted for 19.2% of the whole OTUs, and sequencesof unclassified Eukaryota accounted for 3.3% of thewhole sequences. For bacterial communities, OTU se-quences were taxonomically classified by runningUSEARCH v11 against the Greengenes database [68].All archaea, chloroplasts, eukaryota, mitochondria, andunknown sequences were discarded. Bacterial OTUs

present in < 5 samples with < 10 sequences were re-moved. Finally, the total dataset was randomly normal-ized to 52,248 sequences for each of 116 samples fromstation G, and these sequences were clustered into 16,153 OTUs at a 97% similarity threshold.Considering that the microeukaryotic OTUs identified

at the 97% similarity level with the unoise3 algorithm isnot a specific and accurate estimation of the species orstrain level diversity, we further defined ASVs (ampliconsequence variants) using the unoise3 algorithm, as de-scribed previously [69]. Reads were quality-filtered to amaximum expected error threshold of 1.0, and thenunoise3 was performed to identify ASVs using defaultsettings dataset. In this study, ASVs were included onlyfor beta-diversity analysis of microeukaryotic planktonto assess simply whether our results were biased by theOTUs definition approach.

Real-time quantitative PCRReal-time quantitative PCR (qPCR) was used to quantifythe number of microeukaryotic plankton 18S rRNA genecopies using a LightCycler 480 instrument (Roche, Basel,Switzerland). The 20-μL reaction mixture consisted of10 μL 2 × LightCycler 480 SYBR Green I Master Mix(Roche, Basel, Switzerland), 2 μL DNA template, 0.8 μMof each primer, and 6.4 μL RNase-free water. The PCRruns included tested samples and a negative control intriplicate. The following thermal cycling conditions wereused: 30 s at 94 °C, followed by 40 cycles of 5 s at 94 °C,15 s at 50 °C, and 10 s at 72 °C. Gene fragments were di-luted (108−102 gene copies/μL) to generate the standardcurve using a plasmid containing the 18S rRNA gene.The amplification efficiency (E) of qPCR was calculatedusing the equation E = [10(−1/slope) − 1]. The qPCR effi-ciency of 18S rRNA ranged from 85 to 108% in thisstudy. In addition, qPCR amplification of bacterial 16SrRNA gene was also performed using a Lightcycler 480instrument (Roche, Basel, Switzerland) according to ourprevious method [70]. The qPCR efficiency of 16S rRNAgene was between 95 and 105%.

Definition of core and satellite taxaPartitioning microbial communities into core and satel-lite taxa according to their abundance and occurrencefrequency has contributed to our understanding of com-munity assembly and functioning in many spatiotempo-ral datasets [32, 71]. We arbitrarily defined “core” and“satellite” taxa based on previous study [33]; thus, coretaxa were defined as the OTUs with an occurrence fre-quency ≥ 75% in all samples and satellite taxa as theOTUs with an occurrence frequency < 50% in all sam-ples. Detailed descriptions of core and satellite datasetsare presented in Additional file 1: Figure S4 and Add-itional file 1: Table S2.

Mo et al. Microbiome (2021) 9:128 Page 12 of 17

Statistical analysesAlpha-diversity index (i.e., Shannon-Wiener index) andTukey’s honestly significant difference (Tukey HSD) posthoc test were conducted using R software (version 3.6.1)[72]. The alpha-diversity index was calculated for eachsample using the diversity function in the “vegan” pack-age [72]. Temporal and salinity effects on alpha-diversitywere evaluated using two-way analysis of variance (two-way ANOVA).Microeukaryotic plankton community composition

was visualized using non-metric multidimensional scal-ing (NMDS) based on Bray-Curtis dissimilarities. Ana-lysis of similarity (ANOSIM) was used to evaluatedifferences in microeukaryotic plankton communitiesbetween groups. To reveal the temporal dynamics ofmicroeukaryotic plankton communities, a time-lag re-gression analysis was applied to analyze the Bray-Curtisdissimilarity between each pair of samples, and the timedifference (time-lag) was then plotted against the com-munity dissimilarity. The effects of time and salinity onthe Bray-Curtis dissimilarity were evaluated using Spear-man’s rank correlation. In this study, we used two typesof time: the absolute time span and the annual cycletime span. For example, the time lag between Januaryand December should be (i.e., from January to Decem-ber) 1 rather than 12 in the second time type.In order to study the influence of abiotic (environmen-

tal factors) and biotic (bacterial community) variables onmicroeukaryotic plankton community composition, wecomputed the pairwise Bray-Curtis distances betweensamples on the basis of the relative abundance of micro-eukaryotic plankton (the compositional data). We alsocomputed the pairwise Euclidean distance between thesamples on the basis of environmental data and bacterialalpha- and beta-diversity. Then, partial Mantel test [73]was performed to assess the correlation between micro-eukaryotic plankton community composition and envir-onmental variables or bacterial data, respectively.Further, to determine the relative contribution of time

(in this study: month) and salinity to the assembly ofmicroeukaryotic plankton communities, Mantel and par-tial Mantel tests were applied [73]. The similarity matri-ces of the microeukaryotic plankton community weregenerated based on the Bray-Curtis index. The time andsalinity matrices were obtained using the Euclidean dis-tance. Mantel test assessed the correlation betweenmicroeukaryotic plankton community (Bray-Curtis dis-similarity) and salinity (Euclidean distance) or time (Eu-clidean distance), respectively. The partial Mantel testwas performed to estimate the relative contribution ofsalinity or time variables to the changes in the microeu-karyotic plankton community.In addition, to further study the roles of core and sat-

ellite taxa on ecosystem functions, we calculated the

multi-nutrient cycling index that can track the cycling ofmultiple nutrients in aquatic ecosystem [74] (see Add-itional file 1 for a detailed description). Afterwards, ran-dom forest (RF) machine learning [75] was used toassess the effects of alpha- and beta-diversity of core andsatellite subcommunities on the multi-nutrient cyclingindex (see Additional file 1 for detailed description).

Neutral community modelTo estimate the effects of stochastic processes on themicroeukaryotic plankton community assembly, a neu-tral community model was used [76], applying non-linear least-squares to generate the best fit between thefrequency of OTUs occurrence and their relative abun-dance [77]. R2 value indicates the goodness of fit to themodel, which was calculated following the “Östman’smethod” [78]. When R2 is close to 1, the community as-sembly is fully consistent with stochastic processes.When it does not describe the community composition,R2 can be ≤ 0. Model computations were performed withR version 3.6.1 [72].

Habitat niche breadthTo explore the relative effects of stochastic and deter-ministic processes on microeukaryotic plankton commu-nities, we calculated Levins’ niche breadth (B) index forthe microeukaryotic plankton using the formula:

Bj ¼ 1PN

i¼1Pij2

where Bj indicates the habitat niche breadth of OTU j ina metacommunity; N represents the total number ofcommunities in each metacommunity; Pij is the propor-tion of OTU j in community i [23, 47]. A given OTUwith high B value represents a wide habitat nichebreadth. The community level B value (Bcom) was calcu-lated as the average of B values from all taxa occurringin one given community [23, 49]. A microeukaryoticplankton community with a wide niche breadth is ex-pected to be metabolically more flexible at the commu-nity level than one with a narrow niche breadth [23, 47,49]. The analysis was performed using the “niche.width”function within R package “spaa” [79].

Null modelWe tested clustering or overdispersion of microeukaryo-tic plankton communities by examining the deviation ofeach observed metric from the average of the null model(checkerboard score (C-score)) [80]. The values obtainedwere standardized to allow comparisons among assem-blages using the standardized effect size (SES). Specific-ally, the sequence table was transformed into a binarymatrix of presence (1) and absence (0), and then SES

Mo et al. Microbiome (2021) 9:128 Page 13 of 17

was calculated under the null model [81]. The SES for C-score was estimated as the difference between the observedindex and the mean of the stimulated index over the stand-ard deviation of the stimulated index [82]. The higher orlower SES value than the expected null value is interpretedas overdispersion or underdispersion, respectively, and themagnitude of SES is interpreted as the strength of the effectof deterministic processes on the assemblage [83]. C-scorewas evaluated based on 30,000 simulations and using thesequential swap randomization algorithm with the package“EcoSimR” in R version 3.6.1 [72].

Network constructionWe constructed species co-occurrence networks basedon samples from the whole study period (August 2016–August 2017, 116 samples) for station G in the Xinglin-wan Reservoir. To reduce the complexity of the datasets,we removed OTUs present in less than 20 samples withless than 200 sequences for the construction of net-works. We also constructed community subnetworks forthe all, core, and satellite microeukaryotic planktonbased on samples at the three salinity levels, respectively.SparCC was used to calculate pairwise correlations be-

tween plankton OTUs [84]. Only robust (|r| > 0.6) andstatistically significant (P < 0.01) correlations were incor-porated into the network analyses. Network visualizationwas generated with Cytoscape version 3.6.1 and Gephiversion 0.9.1. Each node indicates a given OTU, andeach edge represents a significant correlation betweentwo OTUs. Degree represents the number of edgesconnecting each node to the rest nodes of the network.Normally, the high topological characteristic values(such as node, edge, and degree) suggest a more com-plex network. In general, there are two common net-work distributions. One is the random network. Arandom network follows a Poisson distribution of edgesper node, meaning that there are no highly associatednodes and that most nodes exhibit a similar number ofedges [85]. The other is non-random network. That is,scale-free or small-world network that has a power-lawdistribution, implying that some nodes are highly associ-ated and maintaining the network together [86, 87].Based on metabolic network approaches [88], the net-

work hubs (Zi-score > 2.5; Pi-score > 0.62), module hubs(Zi-score > 2.5; Pi-score < 0.62), connectors (Zi-score< 2.5; Pi-score > 0.62), and peripherals (Zi-score < 2.5; Pi-score < 0.62) were identified [87]. All hubs and connectornodes could be defined as potential keystone species inco-occurrence networks [89]. Furthermore, the 1000Erdös-Réyni random networks, which exhibit the samenumber of nodes and edges as the real networks, were cal-culated in the “igraph” R package, with each edge havingthe same probability of being assigned to a node [85]. Tofurther describe the topological parameters of the

networks, a set of metrics of both real and random net-works were calculated and compared: clustering coeffi-cient, average path length, and modularity.The network dissimilarity between different salinity

levels was identified using the widely applied equation,which consists of a re-expression of classical measuresof dissimilarity following a partition of shared and totalitems [90, 91]:

βw ¼ aþ bþ c2aþ bþ cð Þ=2−1

where βw is dissimilarity between networks B and C, arepresents number of shared edges between networks Band C, b represents number of edges unique to network B,and c represents number of edges unique to network C.Finally, network stability was evaluated by removing

nodes in the static network to estimate how quickly ro-bustness degraded, and network robustness was assessedby natural connectivity of the nodes. The node removingwas a random repetitive process [92].

Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s40168-021-01079-w.

Additional file 1: Figure S1. Temporal dynamics of 18 majorenvironmental factors in stations C, L, and G from Xinglinwan Reservoirfrom August 12, 2016 to September 20, 2016. Figure S2. Temporaldynamics of 18 major environmental factors in station G from XinglinwanReservoir from August 2016 to August 2017. Figure S3. Changes ofprecipitation and salinity over three years. Figure S4. Definition of core,intermediate, and satellite plankton in the microeukaryoticmetacommunity in Xinglinwan Reservoir. Figure S5. Temporal dynamicof microeukaryotic plankton communities. Figure S6. The absoluteabundance of microeukaryotic plankton and bacterioplankton fromstations C, L, and G in Xinglinwan Reservoir. Figure S7. Variability ofmicroeukaryotic plankton communities across three salinity levels fromthe three stations of Xinglinwan Reservoir. Figure S8. Communitystructuring of microeukaryotic plankton across salinity gradient and timeseries at station G. Figure S9. Comparison of community composition ofall, core and satellite microeukaryotic plankton from station G amongthree salinity levels at phylum level. Figure S10. Opposite changes instochasticity and salinity during the community succession ofmicroeukaryotic plankton. Figure S11. Comparison of mean nichebreadth between core and satellite subcommunities at station G (n =116). Figure S12. The potential importance of core and satellitemicroeukaryotic plankton in community network and nutrient cycle,respectively. Table S1. Previous literatures analyzing the influence ofsalinity changes on plankton or microorganism. Table S2. Description ofmicroeukaryotic plankton OTUs datasets at 97% sequence similarity level.Table S3. Mantel and partial Mantel tests showing the relationshipbetween microeukaryotic plankton community similarity and salinity andtime (month) using Pearson’s coefficient. Table S4. Topologicalproperties of the empirical species co-occurrence networks of microeu-karyotic plankton communities and their associated random network.Table S5. Number of degrees of different microeukaryotic planktongroups in six different modules in integrated networks of station G. TableS6. Numbers of shared OTUs (or nodes) and unique edges (correlations)and their dissimilarity between different microeukaryotic plankton sub-networks based on the three different salinity levels.

Additional file 2.

Mo et al. Microbiome (2021) 9:128 Page 14 of 17

AcknowledgementsThe authors would like to thank Professor David M. Wilkinson, Dr. AnneMette Poulsen, and Dr. Abdullah Al Mamun for constructive comments onan earlier version of this paper.

Authors’ contributionsJ.Y. conceived the idea and designed the research. F.P., X.F.G., K.X.R., and J.Y.collected the samples. Y.Y.M., F.P., and J.Y. performed the experiments. Y.Y.M.,P.X., and J.Y. analyzed the data. Y.Y.M. and J.Y. wrote the first draft of themanuscript, and all authors contributed to and approved the finalmanuscript.

FundingThis work was supported by the Strategic Priority Research Program of theChinese Academy of Sciences (XDA23040302), the National Natural ScienceFoundation of China (91851104, 31672312, 92047204, and 32011530074), andthe Natural Science Foundation of Fujian Province (2019J02016). Theresearch was also supported by the Wuhan Branch, Supercomputing Center,Chinese Academy of Sciences, China. Erik Jeppesen was supported by theTÜBITAK program BIDEB2232 (project 118C250).

Availability of data and materialsAll raw sequences from this study have been stored in the public NCBISequence Read Archive (SRA) database under the BioProject numberPRJNA510458 and the accession number SRP173869 for 18S rRNA gene, andunder BioProject number PRJNA510463 and accession number SRP173857for 16S rRNA gene.

Declarations

Ethics approval and consent to participateNot applicable.

Consent for publicationNot applicable.

Competing interestsThe authors declare that they have no competing interests.

Author details1Aquatic Ecohealth Group, Fujian Key Laboratory of Watershed Ecology, KeyLaboratory of Urban Environment and Health, Institute of UrbanEnvironment, Chinese Academy of Sciences, Xiamen 361021, China.2University of Chinese Academy of Sciences, Beijing 100049, China. 3Instituteof Marine Sciences, CSIC, Passeig Marítim de la Barceloneta 37-49, ES08003Barcelona, Spain. 4Department of Bioscience, Aarhus University, 8600Silkeborg, Denmark. 5Sino-Danish Centre for Education and Research, Beijing100049, China. 6Limnology Laboratory, Department of Biological Sciencesand Centre for Ecosystem Research and Implementation, Middle EastTechnical University, 06800 Ankara, Turkey. 7Institute of Marine Sciences,Middle East Technical University, 33731 Erdemli-Mersin, Turkey.

Received: 24 February 2021 Accepted: 15 April 2021

References1. Brucet S, Boix D, Quintana XD, Jensen E, Nathansen LW, Trochine C, et al.

Factors influencing zooplankton size structure at contrasting temperaturesin coastal shallow lakes: implications for effects of climate change. LimnolOceanogr. 2010;55(4):1697–711. https://doi.org/10.4319/lo.2010.55.4.1697.

2. Dugan HA, Bartlett SL, Burke SM, Doubek JP, Krivak-Tetley FE, Skaff NK, et al.Salting our freshwater lakes. Proc Natl Acad Sci U S A. 2017;114(17):4453–8.https://doi.org/10.1073/pnas.1620211114.

3. Jeppesen E, Beklioğlu M, Özkan K, Akyürek Z. Salinization increase dueto climate change will have substantial negative effects on inlandwaters and freshwater resources: a call for multifaceted research at thelocal and global scale. Innovation. 2020;1(2):100030. https://doi.org/10.1016/j.xinn.2020.100030.

4. Cañedo-Argüelles M, Kefford BJ, Piscart C, Prat N, Schaefer RB, Schulz CJ.Salinisation of rivers: an urgent ecological issue. Environ Pollut. 2013;173:157–67. https://doi.org/10.1016/j.envpol.2012.10.011.

5. Slangen ABA, Church JA, Agosta C, Fettweis X, Marzeion B, Richter K.Anthropogenic forcing dominates global mean sea-level rise since 1970.Nat Clim Chang. 2016;6(7):701–5. https://doi.org/10.1038/nclimate2991.

6. Kaushal SS, Likens GE, Pace ML, Utz RM, Haq S, Gorman J, et al. Freshwatersalinization syndrome on a continental scale. Proc Natl Acad Sci U S A.2018;115(4):E574–83. https://doi.org/10.1073/pnas.1711234115.

7. Corsi SR, Graczyk DJ, Geis SW, Booth NL, Richards KD. A fresh look at roadsalt: aquatic toxicity and water-quality impacts on local, regional, andnational scales. Environ Sci Technol. 2010;44(19):7376–82. https://doi.org/10.1021/es101333u.

8. Hintz WD, Relyea RA. Impacts of road deicing salts on the early-life growthand development of a stream salmonid: salt type matters. Environ Pollut.2017;223:409–15. https://doi.org/10.1016/j.envpol.2017.01.040.

9. Soucek DJ. Sodium sulfate impacts feeding, specific dynamic action, andgrowth rate in the freshwater bivalve Corbicula fluminea. Aquat Toxicol.2007;83(4):315–22. https://doi.org/10.1016/j.aquatox.2007.05.006.

10. Chinathamby K, Reina RD, Bailey PCE, Lees BK. Effects of salinity on the survival,growth and development of tadpoles of the brown tree frog, Litoria ewingii.Aust J Zool. 2006;54(2):97–105. https://doi.org/10.1071/ZO06006.

11. Ji MK, Kong WD, Yue LY, Wang JB, Deng Y, Zhu LP. Salinity reduces bacterialdiversity, but increases network complexity in Tibetan Plateau lakes. FEMSMicrobiol Ecol. 2019;95:fiz190.

12. Simon M, Lopez-Garcia P, Deschamps P, Moreira D, Restoux G, Bertolino P.Marked seasonality and high spatial variability of protist communities in shallowfreshwater systems. ISME J. 2015;9(9):1941–53. https://doi.org/10.1038/ismej.2015.6.

13. Guidi L, Chaffron S, Bittner L, Eveillard D, Larhlimi A, Roux S. Planktonnetworks driving carbon export in the oligotrophic ocean. Nature. 2016;532(7600):465–70. https://doi.org/10.1038/nature16942.

14. Logares R, Lindstrom ES, Langenheder S, Logue JB, Paterson H, Laybourn-Parry J. Biogeography of bacterial communities exposed to progressivelong-term environmental change. ISME J. 2013;7(5):937–48. https://doi.org/10.1038/ismej.2012.168.

15. Nemergut DR, Schmidt SK, Fukami T, O'Neill SP, Bilinski TM, Stanish LF.Patterns and processes of microbial community assembly. Microbiol MolBiol Rev. 2013;77(3):342–56. https://doi.org/10.1128/MMBR.00051-12.

16. Zhou JZ, Ning DL. Stochastic community assembly: does it matter inmicrobial ecology? Microbiol Mol Biol R. 2017;81:e00002–17.

17. Chesson P. Mechanisms of maintenance of species diversity. Annu Rev EcolS. 2000;31(1):343–66. https://doi.org/10.1146/annurev.ecolsys.31.1.343.

18. Fargione J, Brown CS, Tilman D. Community assembly and invasion: anexperimental test of neutral versus niche processes. Proc Natl Acad Sci U SA. 2003;100(15):8916–20. https://doi.org/10.1073/pnas.1033107100.

19. Hubbell SP. The unified neutral theory of biodiversity and biogeography.Princeton: Princeton University Press; 2001.

20. Chave J. Neutral theory and community ecology. Ecol Lett. 2004;7(3):241–53.https://doi.org/10.1111/j.1461-0248.2003.00566.x.

21. Chen WD, Ren KX, Isabwe A, Chen HH, Liu M, Yang J. Stochastic processesshape microeukaryotic community assembly in a subtropical river acrosswet and dry seasons. Microbiome. 2019;7(1):138. https://doi.org/10.1186/s40168-019-0749-8.

22. Logares R, Deutschmann IM, Junger PC, Giner CR, Krabberod AK, SchmidtTSB. Disentangling the mechanisms shaping the surface ocean microbiota.Microbiome. 2020;8(1):55. https://doi.org/10.1186/s40168-020-00827-8.

23. Wu WX, Lu HP, Sastri A, Yeh YC, Gong GC, Chou WC. Contrasting therelative importance of species sorting and dispersal limitation in shapingmarine bacterial versus protist communities. ISME J. 2018;12(2):485–94.https://doi.org/10.1038/ismej.2017.183.

24. Berdjeb L, Parada A, Needham DM, Fuhrman JA. Short-term dynamics andinteractions of marine protist communities during the spring-summer transition.ISME J. 2018;12(8):1907–17. https://doi.org/10.1038/s41396-018-0097-x.

25. Wu WX, Huang BQ. Protist diversity and community assembly in surfacesediments of the South China Sea. MicrobiologyOpen. 2019;8:e891. https://doi.org/10.1002/mbo3.891.

26. Gao C, Montoya L, Xu L, Madera M, Hollingsworth J, Purdom E, et al. Fungalcommunity assembly in drought-stressed sorghum shows stochasticity,selection, and universal ecological dynamics. Nat Commun. 2020;11(1):34.https://doi.org/10.1038/s41467-019-13913-9.

27. Zhang KP, Shi Y, Cui XQ, Yue P, Li KH, Liu XJ, et al. Salinity is a keydeterminant for soil microbial communities in a desert ecosystem.mSystems. 2019;4:e00225–18. https://doi.org/10.1128/mSystems.00225-18.

Mo et al. Microbiome (2021) 9:128 Page 15 of 17

28. Rath KM, Fierer N, Murphy DV, Rousk J. Linking bacterial communitycomposition to soil salinity along environmental gradients. ISME J. 2019;13(3):836–46. https://doi.org/10.1038/s41396-018-0313-8.

29. Tripathi BM, Stegen JC, Kim M, Dong K, Adams JM, Lee YK. Soil pH mediatesthe balance between stochastic and deterministic assembly of bacteria.ISME J. 2018;12(4):1072–83. https://doi.org/10.1038/s41396-018-0082-4.

30. de Vries FT, Griffiths RI, Bailey M, Craig H, Girlanda M, Gweon HS. Soilbacterial networks are less stable under drought than fungal networks. NatCommun. 2018;9(1):3033. https://doi.org/10.1038/s41467-018-05516-7.

31. Magurran AE. Species abundance distributions over time. Ecol Lett. 2007;10(5):347–54. https://doi.org/10.1111/j.1461-0248.2007.01024.x.

32. van der Gast CJ, Walker AW, Stressmann FA, Rogers GB, Scott P, Daniels TW.Partitioning core and satellite taxa from within cystic fibrosis lung bacterialcommunities. ISME J. 2011;5(5):780–91. https://doi.org/10.1038/ismej.2010.175.

33. Magurran AE, Henderson PA. Explaining the excess of rare species in naturalspecies abundance distributions. Nature. 2003;422(6933):714–6. https://doi.org/10.1038/nature01547.

34. Pärtel M, Moora M, Zobel M. Variation in species richness within and betweencalcareous (alvar) grassland stands: the role of core and satellite species. PlantEcol. 2001;157(2):205–13. https://doi.org/10.1023/A:1013938624716.

35. Supp SR, Koons DN, Ernest SKM. Using life history tradeoffs to understandcore-transient structuring of a small mammal community. Ecosphere. 2015;6:1–15. https://doi.org/10.1890/ES15-00239.1.

36. Hanski I. Dynamics of regional distribution: the core and satellite specieshypothesis. Oikos. 1982;38(2):210–21. https://doi.org/10.2307/3544021.

37. Kuiper JJ, van Altena C, de Ruiter PC, van Gerven LPA, Janse JH. Food-webstability signals critical transitions in temperate shallow lakes. Nat Commun.2015;6(1):7727. https://doi.org/10.1038/ncomms8727.

38. Coyte KZ, Schluter J, Foster KR. The ecology of the microbiome: networks,competition, and stability. Science. 2015;350(6261):663–6. https://doi.org/10.1126/science.aad2602.

39. Telesh IV, Khlebovich VV. Principal processes within the estuarine salinitylevel: a review. Mar Pollut Bull. 2010;61(4-6):149–55. https://doi.org/10.1016/j.marpolbul.2010.02.008.

40. Campbell BJ, Kirchman DL. Bacterial diversity, community structure andpotential growth rates along an estuarine salinity level. ISME J. 2013;7(1):210–20. https://doi.org/10.1038/ismej.2012.93.

41. Lu YF, Zhou LH. Eutroprophication evaluation of Xinglin Bay Reservoir. JJimei Univ (Natural Science). 2002;7:295–9.

42. Cañedo-Argüelles M, Kefford B, Schäfer R. Salt in freshwaters: causes, effectsand prospects - introduction to the theme issue. Phil Trans R Soc B. 2019;374(1764):20180002. https://doi.org/10.1098/rstb.2018.0002.

43. Oren A. Thermodynamic limits to microbial life at high salt concentrations.Environ Microbiol. 2011;13(8):1908–23. https://doi.org/10.1111/j.1462-2920.2010.02365.x.

44. Taylor JD, Cunliffe M. Multi-year assessment of coastal planktonic fungireveals environmental drivers of diversity and abundance. ISME J. 2016;10(9):2118–28. https://doi.org/10.1038/ismej.2016.24.

45. Berga M, Zha Y, Szekely AJ, Langenheder S. Functional and compositionalstability of bacterial metacommunities in response to salinity changes. FrontMicrobiol. 2017;8:948. https://doi.org/10.3389/fmicb.2017.00948.

46. Liu LM, Yang J, Lv H, Yu XQ, Wilkinson DM, Yang J. Phytoplanktoncommunities exhibit a stronger response to environmental changes thanbacterioplankton in three subtropical reservoirs. Environ Sci Technol. 2015;49(18):10850–8. https://doi.org/10.1021/acs.est.5b02637.

47. Pandit SN, Kolasa J, Cottenie K. Contrasts between habitat generalists andspecialists: an empirical extension to the basic metacommunity framework.Ecology. 2009;90(8):2253–62. https://doi.org/10.1890/08-0851.1.

48. Dini-Andreote F, Stegen JC, van Elsas JD, Salles JF. Disentanglingmechanisms that mediate the balance between stochastic anddeterministic processes in microbial succession. Proc Natl Acad Sci U S A.2015;112(11):E1326–32. https://doi.org/10.1073/pnas.1414261112.

49. Jiao S, Yang YF, Xu YQ, Zhang J, Lu YH. Balance between communityassembly processes mediates species coexistence in agricultural soilmicrobiomes across eastern China. ISME J. 2020;14(1):202–16. https://doi.org/10.1038/s41396-019-0522-9.

50. Pennekamp F, Pontarp M, Tabi A, Altermatt F, Alther R, Choffat Y.Biodiversity increases and decreases ecosystem stability. Nature. 2018;563(7729):109–12. https://doi.org/10.1038/s41586-018-0627-8.

51. Röttjers L, Faust K. From hairballs to hypotheses-biological insights frommicrobial networks. FEMS Microbiol Rev. 2018;42(6):761–80. https://doi.org/10.1093/femsre/fuy030.

52. Olesen JM, Bascompte J, Dupont YL, Jordano P. The modularity ofpollination networks. Proc Natl Acad Sci U S A. 2007;104(50):19891–6.https://doi.org/10.1073/pnas.0706375104.

53. Ling FQ, Hwang CC, LeChevallier MW, Andersen GL, Liu WT. Core-satellitepopulations and seasonality of water meter biofilms in a metropolitandrinking water distribution system. ISME J. 2016;10(3):582–95. https://doi.org/10.1038/ismej.2015.136.

54. Jiao S, Xu YQ, Zhang J, Hao X, Lu YH. Core microbiota in agricultural soilsand their potential associations with nutrient cycling. mSystems. 2019;4:e00313–8. https://doi.org/10.1128/mSystems.00313-18.

55. Saunders AM, Albertsen M, Vollertsen J, Nielsen PH. The activated sludgeecosystem contains a core community of abundant organisms. ISME J.2016;10(1):11–20. https://doi.org/10.1038/ismej.2015.117.

56. Landi P, Minoarivelo HO, Brannstrom A, Hui C, Dieckmann U. Complexityand stability of ecological networks: a review of the theory. Popul Ecol.2018;60(4):319–45. https://doi.org/10.1007/s10144-018-0628-3.

57. Levins R. Evolution in Changing Environments: Some TheoreticalExplorations (MPB-2). Princeton: Princeton University Press; 1968. https://doi.org/10.1515/9780691209418.

58. Jeppesen ES, Brucet L, Naselli-Flores E, Papastergiadou K, Stefanidis T, NõgesP, et al. Ecological impacts of global warming and water abstraction onlakes and reservoirs due to changes in water level and salinity.Hydrobiologia. 2015;570:201–27. https://doi.org/10.1007/s10750-014-2169-x.

59. Liu LM, Yang J, Yu XQ, Chen GJ, Yu Z. Patterns in the composition ofmicrobial communities from a subtropical river: effects of environmental,spatial and temporal factors. PLoS One. 2013;8(11):e81232. https://doi.org/10.1371/journal.pone.0081232.

60. Amaral-Zettler LA, McCliment EA, Ducklow HW, Huse SM. A method forstudying protistan diversity using massively parallel sequencing of V9hypervariable regions of small-subunit ribosomal RNA genes. PLoS One.2009;4(7):e6372. https://doi.org/10.1371/journal.pone.0006372.

61. Fang PJ, Peng F, Gao XF, Xiao P, Yang J. Decoupling the dynamics ofbacterial taxonomy and antibiotic resistance function in a subtropical urbanreservoir as revealed by high-frequency sampling. Front Microbiol. 2019;10:1448. https://doi.org/10.3389/fmicb.2019.01448.

62. Rognes T, Flouri T, Nichols B, Quince C, Mahe F. VSEARCH: a versatile opensource tool for metagenomics. PeerJ. 2016;4:e2584. https://doi.org/10.7717/peerj.2584.

63. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB.Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities.Appl Environ Microbiol. 2009;75(23):7537–41. https://doi.org/10.1128/AEM.01541-09.

64. Edgar RC. Search and clustering orders of magnitude faster than BLAST.Bioinformatics. 2010;26(19):2460–1. https://doi.org/10.1093/bioinformatics/btq461.

65. Guillou L, Bachar D, Audic S, Bass D, Berney C, Bittner L. The ProtistRibosomal Reference database (PR2): a catalog of unicellulareukaryote small sub-unit rRNA sequences with curated taxonomy.Nucleic Acids Res. 2013;41(Database issue):D597–604. https://doi.org/10.1093/nar/gks1160.

66. Adl SM, Simpson AGB, Lane CE, Lukes J, Bass D, Bowser SS, et al. The revisedclassification of eukaryotes. J Eukaryot Microbiol. 2012;59(5):429–93. https://doi.org/10.1111/j.1550-7408.2012.00644.x.

67. Adl SM, Bass D, Lane CE, Lukes J, Schoch CL, Smirnov A. Revisions to theclassification, nomenclature, and diversity of eukaryotes. J EukaryotMicrobiol. 2019;66(1):4–119. https://doi.org/10.1111/jeu.12691.

68. DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, Keller K, et al.Greengenes, a chimera-checked 16S rRNA gene database and workbenchcompatible with ARB. Appl Environ Microbiol. 2006;72(7):5069–72. https://doi.org/10.1128/AEM.03006-05.

69. Edgar RC. Updating the 97% identity threshold for 16S ribosomal RNAOTUs. Bioinformatics (Oxford, England). 2018;34:2371–5.

70. Nyirabuhoro P, Gao XF, Ndayishimiye JC, Xiao P, Mo YY, Ganjidoust H, et al.Responses of abundant and rare bacterioplankton to temporal change in asubtropical urban reservoir. FEMS Microbiol Ecol. 2021;97:fiab036. https://doi.org/10.1093/femsec/fiab036.

Mo et al. Microbiome (2021) 9:128 Page 16 of 17

71. Ju F, Zhang T. Bacterial assembly and temporal dynamics in activatedsludge of a full-scale municipal wastewater treatment plant. ISME J. 2015;9(3):683–95. https://doi.org/10.1038/ismej.2014.162.

72. R Core Team. R: a language and environment for statistical computing.Vienna: R Foundation for Statistical Computing; 2018.

73. Legendre P, Legendre L. Numerical Ecology. 3rd ed. Netherlands: ElsevierAmsterdam; 2012.

74. Gao XF, Chen HH, Gu B, Jeppesen E, Xue YY, Yang J. Particulate organicmatter as causative factor to eutrophication of subtropical deep freshwater:role of typhoon (tropical cyclone) in the nutrient cycling. Water Res. 2021;188:116470. https://doi.org/10.1016/j.watres.2020.116470.

75. Breiman L. Random forests. Mach Learn. 2001;45(1):5–32. https://doi.org/10.1023/A:1010933404324.

76. Sloan WT, Lunn M, Woodcock S, Head IM, Nee S, Curtis TP. Quantifying theroles of immigration and chance in shaping prokaryote communitystructure. Environ Microbiol. 2006;8(4):732–40. https://doi.org/10.1111/j.1462-2920.2005.00956.x.

77. Elzhov TV, Mullen KM, Spiess AN, Bolker B. Minpack.lm: R interface to theLevenberg-Marquardt nonlinear least-squares algorithm found in MINPACK,plus support for bounds. R package version 1.1-8. 2015. http://CRAN.Rproject.org/package=minpack.lm.

78. Östman Ö, Drakare S, Kritzberg ES, Langenheder S, Logue JB, Lindstrom ES.Regional invariance among microbial communities. Ecol Lett. 2010;13(1):118–27. https://doi.org/10.1111/j.1461-0248.2009.01413.x.

79. Zhang JY. Spaa: Species Association Analysis. R package version 0.2.2. 2016,http://CRAN.R-project.org/package=spaa.

80. Stone L, Roberts A. The checkerboard score and species distributions.Oecologia. 1990;85(1):74–9. https://doi.org/10.1007/BF00317345.

81. Gotelli NJ, Mccabe DJ. Species co-occurrence: a meta-analysis of J. M.Diamond’s assembly rules model. Ecology. 2002;83:2091–6. https://doi.org/10.1890/0012-9658(2002)083[2091:SCOAMA]2.0.CO;2.

82. Crump BC, Peterson BJ, Raymond PA, Amon RMW, Rinehart A, McClelland JW,et al. Circumpolar synchrony in big river bacterioplankton. Proc Natl Acad Sci US A. 2009;106(50):21208–12. https://doi.org/10.1073/pnas.0906149106.

83. Swenson NG. Functional and phylogenetic ecology in R. New York:Springer; 2014. https://doi.org/10.1007/978-1-4614-9542-0.

84. Friedman J, Alm EJ. Inferring correlation networks from genomic surveydata. PLoS Comput Biol. 2012;8(9):e1002687. https://doi.org/10.1371/journal.pcbi.1002687.

85. Erdös P, Rényi A. On the evolution of random graphs. Publ Math InstHungarian Acad Sci. 1960;5:17–61.

86. Banerjee S, Schlaeppi K, van der Heijden MGA. Keystone taxa as drivers ofmicrobiome structure and functioning. Nat Rev Microbiol. 2018;16(9):567–76. https://doi.org/10.1038/s41579-018-0024-1.

87. Strogatz SH. Exploring complex networks. Nature. 2001;410(6825):268–76.https://doi.org/10.1038/35065725.

88. Guimerà R, Amaral LAN. Functional cartography of complex metabolic networks.Nature. 2005;433(7028):895–900. https://doi.org/10.1038/nature03288.

89. Shi SJ, Nuccio EE, Shi ZJ, He ZL, Zhou JZ, Firestone MK. The interconnectedrhizosphere: High network complexity dominates rhizosphere assemblages.Ecol Lett. 2016;19(8):926–36. https://doi.org/10.1111/ele.12630.

90. Koleff P, Gaston KJ, Lennon JJ. Measuring beta diversity for presence-absence data. J Anim Ecol. 2003;72(3):367–82. https://doi.org/10.1046/j.1365-2656.2003.00710.x.

91. Poisot T, Canard E, Mouillot D, Mouquet N, Gravel D. The dissimilarity ofspecies interaction networks. Ecol Lett. 2012;15(12):1353–61. https://doi.org/10.1111/ele.12002.

92. Peng GS, Wu J. Optimal network topology for structural robustness basedon natural connectivity. Physica A. 2016;443:212–20. https://doi.org/10.1016/j.physa.2015.09.023.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Mo et al. Microbiome (2021) 9:128 Page 17 of 17